modify the BRIDGE examples and rect model

examples/bridge/bridge_tacm12k.py

@@ -44,22 +44,20 @@
     _,
 ) = dataset.data_list
 
-# Making graph
-paper_embeddings = paper_embeddings.to(device)
 adj = build_homo_adj(
     relation_df=citations_table.df,
     n_all=len(papers_table),
 ).to(device)
 target_table = papers_table.to(device)
 y = papers_table.y.long().to(device)
+paper_embeddings = paper_embeddings.to(device)
 
 
 train_mask, val_mask, test_mask = (
     papers_table.train_mask,
     papers_table.val_mask,
     papers_table.test_mask,
 )
-out_dim = papers_table.num_classes
 
 
 class Bridge(torch.nn.Module):
@@ -79,6 +77,29 @@ def forward(self, table, non_table, adj):
         return node_feats[: len(table), :]
 
 
+t_encoder = TableEncoder(
+    in_dim=paper_embeddings.size(1),
+    out_dim=paper_embeddings.size(1),
+    table_transorm=FTTransformerTransform(col_stats_dict=papers_table.stats_dict),
+    table_conv=TabTransformerConv,
+)
+g_encoder = GraphEncoder(
+    in_dim=paper_embeddings.size(1),
+    out_dim=papers_table.num_classes,
+    graph_transform=GCNNorm(),
+    graph_conv=GCNConv,
+)
+model = Bridge(
+    table_encoder=t_encoder,
+    graph_encoder=g_encoder,
+).to(device)
+optimizer = torch.optim.Adam(
+    model.parameters(),
+    lr=args.lr,
+    weight_decay=args.wd,
+)
+
+
 def accuracy_score(preds, truth):
     return (preds == truth).sum(dim=0) / len(truth)
 
@@ -112,32 +133,8 @@ def test_epoch():
     return train_acc.item(), val_acc.item(), test_acc.item()
 
 
-t_encoder = TableEncoder(
-    in_dim=paper_embeddings.size(1),
-    out_dim=paper_embeddings.size(1),
-    table_transorm=FTTransformerTransform(col_stats_dict=papers_table.stats_dict),
-    table_conv=TabTransformerConv,
-)
-g_encoder = GraphEncoder(
-    in_dim=paper_embeddings.size(1),
-    out_dim=out_dim,
-    graph_transform=GCNNorm(),
-    graph_conv=GCNConv,
-)
-model = Bridge(
-    table_encoder=t_encoder,
-    graph_encoder=g_encoder,
-).to(device)
-
 start_time = time.time()
 best_val_acc = best_test_acc = 0
-optimizer = torch.optim.Adam(
-    [
-        dict(params=model.table_encoder.parameters(), lr=0.001),
-        dict(params=model.graph_encoder.parameters(), lr=0.01, weight_decay=1e-4),
-    ]
-)
-
 for epoch in range(1, args.epochs + 1):
     train_loss = train_epoch()
     train_acc, val_acc, test_acc = test_epoch()

examples/bridge/bridge_tlf2k.py

@@ -59,7 +59,6 @@
     artist_table.val_mask,
     artist_table.test_mask,
 )
-out_dim = artist_table.num_classes
 
 
 class Bridge(torch.nn.Module):
@@ -79,6 +78,29 @@ def forward(self, table, non_table, adj):
         return node_feats[: len(table), :]
 
 
+t_encoder = TableEncoder(
+    in_dim=emb_size,
+    out_dim=emb_size,
+    table_transorm=FTTransformerTransform(col_stats_dict=artist_table.stats_dict),
+    table_conv=TabTransformerConv,
+)
+g_encoder = GraphEncoder(
+    in_dim=emb_size,
+    out_dim=artist_table.num_classes,
+    graph_transform=GCNNorm(),
+    graph_conv=GCNConv,
+)
+model = Bridge(
+    table_encoder=t_encoder,
+    graph_encoder=g_encoder,
+).to(device)
+optimizer = torch.optim.Adam(
+    model.parameters(),
+    lr=args.lr,
+    weight_decay=args.wd,
+)
+
+
 def accuracy_score(preds, truth):
     return (preds == truth).sum(dim=0) / len(truth)
 
@@ -112,27 +134,8 @@ def test_epoch():
     return train_acc.item(), val_acc.item(), test_acc.item()
 
 
-t_encoder = TableEncoder(
-    in_dim=emb_size,
-    out_dim=emb_size,
-    table_transorm=FTTransformerTransform(col_stats_dict=artist_table.stats_dict),
-    table_conv=TabTransformerConv,
-)
-g_encoder = GraphEncoder(
-    in_dim=emb_size,
-    out_dim=out_dim,
-    graph_transform=GCNNorm(),
-    graph_conv=GCNConv,
-)
-model = Bridge(
-    table_encoder=t_encoder,
-    graph_encoder=g_encoder,
-).to(device)
-
-
 start_time = time.time()
 best_val_acc = best_test_acc = 0
-optimizer = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.wd)
 for epoch in range(1, args.epochs + 1):
     train_loss = train_epoch()
     train_acc, val_acc, test_acc = test_epoch()

examples/bridge/bridge_tml1m.py

@@ -63,7 +63,6 @@
     user_table.val_mask,
     user_table.test_mask,
 )
-out_dim = user_table.num_classes
 
 
 class Bridge(torch.nn.Module):
@@ -83,6 +82,29 @@ def forward(self, table, non_table, adj):
         return node_feats[: len(table), :]
 
 
+t_encoder = TableEncoder(
+    in_dim=emb_size,
+    out_dim=emb_size,
+    table_transorm=FTTransformerTransform(col_stats_dict=user_table.stats_dict),
+    table_conv=TabTransformerConv,
+)
+g_encoder = GraphEncoder(
+    in_dim=emb_size,
+    out_dim=user_table.num_classes,
+    graph_transform=GCNNorm(),
+    graph_conv=GCNConv,
+)
+model = Bridge(
+    table_encoder=t_encoder,
+    graph_encoder=g_encoder,
+).to(device)
+optimizer = torch.optim.Adam(
+    model.parameters(),
+    lr=args.lr,
+    weight_decay=args.wd,
+)
+
+
 def accuracy_score(preds, truth):
     return (preds == truth).sum(dim=0) / len(truth)
 
@@ -116,27 +138,8 @@ def test_epoch():
     return train_acc.item(), val_acc.item(), test_acc.item()
 
 
-t_encoder = TableEncoder(
-    in_dim=emb_size,
-    out_dim=emb_size,
-    table_transorm=FTTransformerTransform(col_stats_dict=user_table.stats_dict),
-    table_conv=TabTransformerConv,
-)
-g_encoder = GraphEncoder(
-    in_dim=emb_size,
-    out_dim=out_dim,
-    graph_transform=GCNNorm(),
-    graph_conv=GCNConv,
-)
-model = Bridge(
-    table_encoder=t_encoder,
-    graph_encoder=g_encoder,
-).to(device)
-
-
 start_time = time.time()
 best_val_acc = best_test_acc = 0
-optimizer = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.wd)
 for epoch in range(1, args.epochs + 1):
     train_loss = train_epoch()
     train_acc, val_acc, test_acc = test_epoch()

examples/rect.py

@@ -11,7 +11,7 @@
 
 # Datasets              Citeseer    |         Cora         | Pubmed
 # Unseen Classes  [1, 2, 5]  [3, 4] | [1, 2, 3]  [3, 4, 6] | [2]
-# RECT-L          66.50      68.40  | 74.80      72.20     | 75.30
+# RECT-L          61.10      66.10  | 71.20      70.40     | 69.90
 
 import argparse
 import copy
@@ -41,7 +41,7 @@
     [
         UT.NormalizeFeatures("l2"),
         UT.SVDFeatureReduction(200),
-        GT.GDC(),
+        GT.GCNNorm(),
     ]
 )
 

rllm/nn/models/rect.py

@@ -30,6 +30,7 @@ def __init__(
         self.in_dim = in_dim
         self.hidden_dim = hidden_dim
         self.dropout = dropout
+        self.prelu = torch.nn.PReLU()
         self.conv = GCNConv(in_dim, hidden_dim)
         self.lin = Linear(hidden_dim, in_dim)
         self.reset_parameters()
@@ -41,7 +42,7 @@ def reset_parameters(self):
         torch.nn.init.xavier_uniform_(self.lin.weight.data)
 
     def forward(self, x: Tensor, adj: Tensor):
-        x = self.conv(x, adj)
+        x = self.prelu(self.conv(x, adj))
         x = F.dropout(x, p=self.dropout, training=self.training)
         return self.lin(x)